non-representative sample. you can clearly see that there are way more dots in the 80~90 range than there are in the 90~110 range. if the sample was representative, the mean would be around 100. with a lower mean (here it seems to be around 80), naturally, there will be more dots of lower IQ with average income, and the plot will give the impression that IQ and income are independent variables.
that's like measuring height and basketball performance, but then you assess the basketball performance of 20000 short people and 500 tall people, plot the graph, cover the lower end of the basketball performance with a rectangle, and now it looks like there are as many short people who are average or good at basketball as there are tall people in the same performance range.
